Re: Questions about the etymology model

[John McCrae <john.mccrae@insight-centre.org>]

Hi Michal,

On 07/11/2022 12:03, Michal MÄchura wrote:

Hi John,

I have three questions about the etymology module.

Question 1

Is the etymology object type, as a child of entry, necessary? It seems like itâs just a container for a list of etymons (and multi-etymons, and etym descriptions). If itâs just a container for things, then why not get rid of it and attach those things to entry directly? Thatâs how we do it elsewhere in the standard, for example, we donât have a container for senses, we attach each sense directly as a child of entry.

The only reason why etymology might be necessary is if you want to make it possible for dictionary authors to have more than one in a single entry, where eaxch etymology object would represent a different hypothesis about the headwordâs etymology. If this is how you meant it, then fine, but maybe we should say so in the text.

Yes, I think the point is that there may be a chain of etymons (e.g., Middle English < Old English < Latin < PIE) but there may also be alternative proposed etymologies. Multiple etymons in the same etymology represent a chain of inheritances, whereas multiple etymologies represent alternative histories.

This should be documented, of course.

Question 2

Why is etymon allowed to have more than one etymDate, that is, more than one âfrom-untilâ timespan? If an etymon represents one step in the etymological evolution of the headword, then I would imagine that the step lasted for exactly one timespan, not several.

    This is probably not necessary. I will make it a 0..1 property

Question 3

Etymons can be grouped into multi-etymons (the multiEtymon object type) to indicate that the headword was derived by combining two etymons. Thatâs fine and makes sense. But then the object model as you have designed it would allow for each etymon inside a multi-etymon to have different timespans (because time is assigned to etymons, not to multi-etymons). This seems unsafe to me, it fails to prohibit a contradiction.

I would propose an alternative object model where we distinguish between timespans and etymons. The etymology of a headword would be a list of timespans, and each timespan would have one or more etymons, where the intended semantics is that if it has more than one, then the etymons were compounded to form the headword:

- entry
    - etymology
        - timespan (from: ... until: ...)
            - etymon: ...
        - timespan (from: ... until: ...)
            - etymon: ...
        - timespan (from: ... until: ...)
            - etymon: ...
            - etymon: ...

It is not an unreasonable modelling and a good point. I would be concerned about making timespan mandatory as it is rarely specified in ontologies. I think it would be cleaner to document that multi-etymons have a date and that etymons under a multi-etymon should not have a date. We can discuss this on Monday.


Regards,

John